NVIDIA has introduced Cosmos 3, an open world foundation model designed to accelerate the development of physical AI systems such as robots, autonomous vehicles, and vision-based AI agents. Built on a new mixture-of-transformers architecture, the model combines vision reasoning, world simulation, and action prediction within a single framework.
Cosmos 3 can understand and generate text, images, videos, sounds, and actions, helping developers create synthetic training data and build AI systems with less data and lower training costs. NVIDIA says the model achieved leading results across several physical AI benchmarks covering world generation, action policies, and visual understanding.
こちらもお読みください: AIボイスが企業のコミュニケーションツールを変革
Alongside the launch, NVIDIA sort of announced the Cosmos Coalition, which is an ecosystem of AI and robotics companies including Runway, Skild AI, Agile Robots and Black Forest Labs. The whole initiative is meant to back open development of foundational AI models and, in the same breath, speed up novel progress in robotics, self-driving, and industrial AI use cases.


